feat: migrate on-device Qwen3 ASR to CoreML for background Flow dictation

Replace MLX GPU inference with CoreML bundles so transcription continues while the host app is backgrounded. Adds model download and warm-up, vendored Qwen3Speech, and updates onboarding, settings, and copy for the ~1.6 GB CoreML package (iOS 18+).
2026-06-23 00:46:58 +08:00
parent 5e5122f172
commit df1c5ff32c
160 changed files with 22080 additions and 492 deletions
@@ -29,7 +29,7 @@ Open an issue using the **Feature request** template. Briefly describe:
 2. **Generate the project locally:**
   ```bash
   brew install xcodegen swiftlint
-   xcodegen generate
+   xcodegen generate          # or: ./Scripts/generate-xcodeproj.sh
   ```
 3. **Code style.** SwiftLint config lives in `.swiftlint.yml` — keep it green. We use Swift 6 strict concurrency, no `Sendable` shims where avoidable.
 4. **Tests.** Add XCTest coverage in `OSGKeyboardTests/` for any non-trivial logic.